Japan: Police Made First Arrest Over Sexual AI Deepfake Images
Tokyo police have made the first arrest in Japan over the use of generative artificial intelligence to create and sell sexually explicit deepfake images of celebrities.
The Metropolitan Police Department on Oct. 16 said Hiroya Yokoi, a 31-year-old office worker from Akita, was arrested on suspicion of producing and displaying obscene materials, specifically thousands of fake images of more than 260 female celebrities.
Police said Yokoi has admitted to the allegations.
“The reaction from viewers was huge, and it became so popular that I realized I could make a lot of money”, police quoted the suspect as saying.
According to police, the case is the first of its kind in Japan.
The victims included more than 260 television personalities, actors, idols and news announcers.
According to police, from January to June this year, Yokoi created and posted explicit images modeled on three female celebrities to an online communication site, making them available for anyone to view.
Yokoi reportedly told investigators he began the operation in October last year to earn extra cash for living expenses and student loan payments after seeing others making money the same way.
AI PORN AS SIDE HUSTLE
Police suspect Yokoi used a promotional account on social media to funnel users to a subscription site where he displayed about 20,000 deepfake images. Subscribers could request custom images for a premium fee.
At least 50 people subscribed, and in the 11 months since October, the scheme netted him an estimated 1.2 million yen ($8,000), investigative sources said.
According to the sources, the suspect created the images as a “side job”.
He had no professional training in generative AI or information and communication technology, having taught himself through online articles and videos.
The free software he used has advertised its ability to “generate high-quality images in seconds.”
Yokoi trained the AI with celebrity photos and used specific text “prompts” to generate the expansive library of images, the sources said.
GLOBAL TRENDS, LEGAL LOOPHOLES
This ease of use has contributed to a global surge in sexual deepfakes.
Research from a U.S. security firm identified 95,820 deepfake videos online in 2023 — a 5.5-fold increase from four years ago — with 98 percent being of a sexual nature.
In response, the United States, Britain and other countries have been strengthening regulations.
South Korea toughened its penalties last September, making not only the creation but also the possession and viewing of sexual deepfakes punishable offenses.
But Japan currently has no specific laws to regulate sexual deepfakes, forcing authorities to operate within the scope of existing legislation.
Japanese law enforcement officers must overcome the high hurdle of obtaining objective evidence that proves an AI-generated image “legally possesses” the identity of the person it depicts.
Opening a defamation investigation requires a complaint from the victim, who is often unaware of the deepfake images. By the time they realize they have been unfairly depicted, the images have already spread widely.
“The creators may be doing this casually, but the victims suffer damage to their reputations and can be harmed mentally and financially,” a Japanese investigator said. “We have to put the brakes on businesses that misuse generative AI”.



Post Comment